Professional Documents
Culture Documents
Nicola Belle
Paola Cantarelli
Bocconi University, Italy
A
dherence to the highest standards of Despite the importance of ethics in government,
ethical conduct is inherent in the mission which remains “at the heart of what we are about
of government organizations, the purpose as a professional field” (Perry 2015, 187), public
of which is to serve the public interest. Indeed, a administration research in this area has been scant
long-standing tradition in the public administration (for a review, see Menzel 2015). Adams and Balfour’s
literature suggests that ethics and morality are search for “scholarly articles on ethics topics in leading
two founding principles of civil service. Using public administration journals indicates that, at
normative approaches, scholars have specified their zenith in the last two decades of the twentieth
what civil servants should do and how they century, ethics topics hovered around 5 percent of
should behave (e.g., Cooper 2012; Perry 2015; all articles. In addition, their topical coverage is slim”
Pugh 1991; Rainey 2009; Rohr 1986, 1998). (2010b, 781). Also, a recent virtual issue of Public
The members of the American Society for Public Administration Review containing selected studies
Administration (ASPA) have committed themselves that the journal has published since 2002 on the
to the demonstration of personal integrity and the topics of “corruption, unethical behavior, and ethics,”
promotion of ethical organizations, two principles included only 13 articles. In short, “studies addressing
in the Code of Ethics approved by the ASPA this topic are needed, important, and long overdue”
National Council in 2013 (ASPA 2016; Svara (Adams and Balfour 2010b, 783). The paucity of
2014). Nevertheless, daily news uncovers major public administration scholarship on ethics contrasts
and minor acts of dishonesty by public officials. with a spurt in other disciplines within the social
The relevance of this concern is demonstrated sciences, such as business, economics, management,
by the fact that one stream of research in public and psychology, in which the investigation of the
Public Administration Review,
administration is focused on investigating determinants of unethical behavior has drawn Vol. 77, Iss. 3, pp. 327–339. © 2017 by
administrative evil and how it threatens traditional increasing attention among scholars in recent years The American Society for Public Administration.
public values (e.g., Adams and Balfour 2015). (e.g., Moore and Gino 2015). DOI: 10.1111/puar.12714.
What Causes Unethical Behavior? A Meta-Analysis to Set an Agenda for Public Administration Research 327
The public administration research on ethics is characterized by from other disciplines is to inspire public administration scholars
a methodological delay relative to other fields. To the best of our to conduct randomized controlled trials on samples of public
knowledge, unlike the evidence available in public administration administration students and employees, which may help nurture
journals to date (for reviews, see Menzel 2015; Von Maravić 2008), the resurgence of public ethics research in our own field. Indeed,
it has become standard practice in other disciplines to investigate Adams and Balfour talk about “the complicity of academic research
the causes of unethical behavior using randomized experiments in perpetuating the ethical theory/ethical practice disconnection by
(e.g., Moore and Gino 2015). Because the public administration marginalizing the study of public service ethics” (2010b, 767).
literature on unethical behavior has relied almost exclusively on
observational designs, our field still lacks a solid understanding of Because almost a decade has passed since the last research synthesis
the causal mechanisms underlying unethical behavior in public on the determinants of unethical conduct (Kish-Gephart, Harrison,
organizations. and Treviño 2010), and experimental research has grown fast,
we believe the time is ripe for a comprehensive meta-analysis of
The primary aim of our work is to help fill this gap by providing experimental literature on the causes of unethical behavior. We are
a comprehensive synthesis of the experimental literature in this convinced that our study substantially advances previous meta-
area. We collected data from 137 experiments in 73 articles and analytic work on the same topic. In particular, we depart from
identified 12 causes of unethical behavior. To address brevity Kish-Gephart, Harrison, and Treviño (2010) in several important
concerns, this article report results in full detail for the seven factors ways. For instance, those authors analyzed only two out of the 137
(social influences, greed, egocentrism, monitoring, moral reminders, experiments included in our sample. This lack of overlap is not
self-justification, and self-view) for which we meta-analyzed more surprising if we consider that the spurt of experimental scholarship
than 10 primary experiments. Findings for the remaining five on the determinants of unethical conduct is a very recent
factors (self-control depletion, slippery slope effect, loss aversion, phenomenon. Indeed, 86 percent of the experiments on unethical
goal setting, and time pressure) are summarized in table 1 and made behavior that we meta-analyze were published after 2008, which
available in the online supplement (tables S8–S12 and table S15). is the last year covered by Kish-Gephart, Harrison, and Treviño’s
To identify the factors for which we report fully detailed analyses, meta-analysis. The novelty of our meta-analytic synthesis is also
we used the 10-study cutoff that is usually suggested as a threshold demonstrated by a striking discrepancy between the determinants
for conducting meaningful publication bias analyses (e.g., Banks, of unethical behavior identified by Kish-Gephart, Harrison, and
Kepes, and Banks 2012; Cantarelli, Belardinelli, and Belle 2016; Treviño and those resulting from our comprehensive meta-analysis
Kepes et al. 2012). of experimental work on this topic.
We limited our meta-analyses to experimental research for two main Ethics Literature and Research in Public Administration
reasons. First, we are interested in investigating the causal mechanisms Normative debates about public administration ethics focus on two
that drive unethical behavior; randomized trials best serve this sets of administrative values: bureaucratic ethos and democratic
purpose because “no other scientific method regularly matches the ethos (Pugh 1991). One key element of the former is that in
characteristics of causal relationships so well” (Shadish, Cook, and providing public services, civil servants have to consider managerial
Campbell 2002, 7). Second, we hope to motivate scholars in our values such as efficiency, effectiveness, loyalty, and accountability
field to conduct future experimental research on the determinants (Pugh 1991). Without an explicit reference to public ethics values,
of unethical behavior that will capitalize on what we know now, and with the prominence of technical rationality, bureaucratic ethos
complement normative debates, and triangulate observational can turn into administrative evil (Adams 2011; Adams and Balfour
findings. The ultimate goal of presenting experimental meta-evidence 2010a, 2015; Zanetti and Adams 2000), under which “ordinary
Notes: Data from one experiment can be included in more than one factor. Factors marked with the asterisk indicate meta-analyses for which we do not provide fully
detailed results in-text but we provided all analyses and information in the online supplement (Tables S8–S12 and Table S15).
Lastly, we conducted a subgroup analysis to test for heterogeneity We also conducted a subgroup analysis to test for heterogeneity
across types of outcome observed in primary studies. Hedges’s g across experimental settings. Hedges’s g point estimates of the
point estimates of the effect sizes were .50 (p < .001, N = 25) for effect sizes were .52 (p < .001, N = 16) for experiments conducted
experiments measuring actual behavior, .41 (p < .001, N = 4) for in laboratory settings and .53 (p < .001, N = 2) for online
experiments measuring intentions to behave, and .47 (p = .009, experiments (Q-value = .002, p = .966). Thus, we found no evidence
N = 3) for experiments measuring judgment of the degree of of heterogeneity of the summary effect sizes across experimental
unethicality (Q-value = .79, p = .852). Therefore, we found no settings. Publication bias analyses returned mixed results (table
evidence of heterogeneity of effects sizes across types of outcomes S14).
being measured in original studies. Evidence of publication bias was
mixed (table S14). Egocentrism
Self-construal theory suggests that one way in which individuals
Greed define themselves is through their unique traits and independence
Recent experimental work drawing on distributive justice (Deutsch from others. In the state of independent self, individuals are less
1985), equity theory (Adams 1965), and social comparison likely to update their self-concept when acting unethically and less
processes (Festinger 1954; Wheeler, Koestner, and Driver 1982) likely to consider the degree to which their behaviors violate social
suggests that individuals are more likely to engage in unethical norms and/or affect others (e.g., Cojuharenco et al. 2011; Sedikides
behavior if they perceive that they are treated unfairly relative to and Brewer 2001). Along the same lines, social value orientation
their peers (e.g., Gino and Pierce 2011; Houser, Vetter, and Winter (Messick and McClintock 1968; Van Lange 1999) argues that
2012; Yam, Reynolds, and Hirsh 2014). Individuals compare pro-self-oriented individuals care about maximizing their benefits
their input-to-output ratio with a referent other around them, and are not concerned about the outcomes they cause onto others.
and whenever perceptions of inequity arise, they will tend to take Thus, they are expected to have less stringent moral standards and
actions to restore equity (Adams 1965). This includes engaging to be more likely to undertake an unethical conduct (Folmer and
in unethical behavior (Gino and Pierce De Cremer 2012). Interdependence theory
2009, 2011). Scholars have suggested that Individuals may become greedy also claims that the desire to maximize one’s
perceptions of inequity may be triggered and engage in unethical behav- own outcome allows individuals not to update
by the mere presence of abundant wealth their self-concept when behaving unethically
available in the environment, even in the
ior to restore fairness in the (Insko et al. 1990; Wiltermuth 2011).
absence of differentials in the input-to-output presence of both pay differen-
ratio (Gino and Pierce 2009). In other words, tials and wealth abundance in We retrieved 20 experiments testing whether
individuals may become greedy and engage the environment. individuals who perceive themselves as
in unethical behavior to restore fairness in the independent rather than related to others
presence of both pay differentials and wealth tend to be more dishonest. The first type of
abundance in the environment. experiments in this meta-analysis tested whether unethical behavior
increases with competition (e.g., Rigdon and D’Esterre 2015). The
We identified 26 experiments testing the effects of greed on second type of studies investigated whether the conflict of interest
individuals’ unethical behavior. The first type of experiments in causes higher levels of unethicality (e.g., Sah and Loewenstein
this meta-analysis tested the effects of inducing individuals to 2015). The last type of experiments tested whether individuals who
think about money or exposing them to wealth abundance in the are more focused on their own self-interest tend to behave more
environment on their unethicality (e.g., Gino and Mogilner 2014). dishonestly (e.g., Winterich, Mittal, and Morales 2014).
The second type of studies investigated the causal effect of offering
individuals a monetary incentive on their levels of dishonesty The results from the meta-analysis of the 20 independent
(e.g., Sharma et al. 2014). The third type of studies tested whether experiments (n = 2,770) showed a significant effect of egocentrism
individuals are more likely to engage in unethical behavior when on unethicality. The Hedges’s g was .58 (p < .001) (table S3). The
they are worse off relative to their peers (e.g., John, Loewenstein, heterogeneity test indicated significant variability among studies
and Rick 2014). (Q-value (19) = 51.54 (p < .001); I2 = 63.13 percent; T2 = .054).
We conducted a subgroup analysis to test for heterogeneity across
The results from the meta-analysis of the 26 independent disciplines. Hedges’s g point estimates of the effect sizes were .83
experiments (n = 2,776) showed a significant effect of greed on (p < .001, N = 3) for business, .27 (p = .082, N = 3) for economics,
individuals’ unethicality. The Hedges’s g was.45 (p < .001) (table and .60 (p < .001, N = 11) for psychology (Q-value = 4.80, p = .091).
S2). The heterogeneity test indicated significant variability Therefore, we failed to reject at the .05 level but not at the .10 level
among studies (Q-value (25) = 139.00 (p < .001); I2 = 82.02 the null hypothesis that effect sizes were the same across disciplines.
percent; T2 = .164). We conducted a subgroup analysis to test for
heterogeneity across disciplines. Hedges’s g point estimates of the We also conducted a subgroup analysis to test for heterogeneity
effect sizes were .28 (p = .574, N = 2) for business, .20 (p = .342, across experimental settings. Hedges’s g point estimates of the
N = 7) for economics, .34 (p = .001, N = 2) for operations, and effect sizes were .48 (p < .001, N = 10) for experiments conducted in
332 Public Administration Review • May | June 2017
laboratory settings and .64 (p < .001, N = 8) for online experiments analysis exposed participants either to a code of ethics (e.g., Mazar,
(Q-value = 1.65, p = .199). Thus, effect sizes did not seem to vary Amir, and Ariely 2008) or to ethical priming exercises (e.g., Welsh
by experimental settings. Lastly, we ran a subgroup analysis by type and Ordóñez 2014).
of participants. Hedges’s g point estimates of the effect sizes were
.87 (p < .001, N = 3) for alumni, .48 (p < .001, N = 4) for the general Meta-analytic results showed a significant effect of moral reminders
population, and .46 (p < .001, N = 11) for students (Q-value = 19.09, on individuals’ unethical behavior. The Hedges’s g was –.43
p < .001). Thus, effect sizes significantly varied across types of (p < .001) (table S5). The heterogeneity test indicated that the
participants. Measures of publication bias suggested that it is variability among studies is not significant at the .05 level (Q-value
unlikely (table S14). (14) = 21.66 (p = .086); I2 = 35.35 percent; T2 = .023). We conducted
a subgroup analysis to test for heterogeneity of summary effect sized
Monitoring across disciplines. Hedges’s g point estimates of the effect sizes were
The extant literature has suggested that monitoring individuals’ –.37 (p < .001, N = 8) for business and –.43 (p = .003, N = 4) for
behavior reduces unethicality in two ways. On the one hand, psychology (Q-value = .16, p = .694). Therefore, there is no evidence
monitoring decreases the perceptions that unethical acts go of heterogeneity of summary effect sizes across disciplines.
undetected (Welsh and Ordóñez 2014) and that the unethical actor
will not be held responsible (Mazar and Aggarwal 2011). On the We also conducted a subgroup analysis to test for heterogeneity
other hand, monitoring draws individuals’ attention to their moral across experimental settings. Hedges’s g point estimates of the effect
standards and encourages self-awareness. In particular, individuals sizes were –.54 (p < .001, N = 10) for experiments conducted in
who are aware of being watched are less likely to engage in unethical laboratory settings and –.34 (p = .002, N = 2) for online experiments
actions without updating their self-concept (Alge, Greenberg, (Q-value = 2.08, p = .149). Thus, effect sizes did not seem to vary
and Brinsfield 2006; Welsh and Ordóñez 2014). We identified by experimental settings. Also, we ran subgroup analyses by type
19 experiments that tested the effects of monitoring individuals’ of participants. Hedges’s g point estimates of the effect sizes were
behavior on their levels of unethicality. Most experiments in this –.34 (p = .002, N = 2) for the general population, –.47 (p < .001,
meta-analysis investigated whether individuals are more likely to N = 7) for students, –.72 (p < .001, N = 4) for students and university
engage in unethical behavior when they are not being monitored employees, and –.07 (p = .58, N = 2) for workers (Q-value = 13.87,
compared with when they are controlled (e.g., Rixom and Mishra p = .003). Thus, we found strong evidence that effect sizes differed
2014). The remaining studies compared unethical behavior under by type of participants.
higher or lower degree of visibility of the dishonest actor or action
(e.g., Ploner and Regner 2013). Lastly, we ran a subgroup analysis by observed outcome. Hedges’s
g point estimates of the effect sizes were –.45 (p < .001, N = 13) for
The results from the meta-analysis of the 19 independent actual behavior and –.34 (p = .002, N = 2) for intentions to behave
experiments (n = 2,575) showed a negative effect of monitoring (Q-value = .59, p = .444). Thus, effect sizes did not vary by outcomes
on individuals’ unethicality. The Hedges’s g was –.84 (p < .001) measured in primary studies. Measures of publication bias suggest
(table S4). The heterogeneity test indicated significant variability that it is unlikely (table S14).
among studies (Q-value (18) = 178.03 (p < .001); I2 = 89.89 percent;
T2 = .32). We conducted a subgroup analysis to test for heterogeneity Self-Justification
of summary effect sizes across disciplines. Hedges’s g point estimates Social cognitive theory (Bandura 1986) suggests that the availability
of the effect sizes were –.61 (p < .001, N = 6) for business, and of justifications encourages unethical behavior through an increased
–.99 (p < .001, N = 12) for psychology (Q-value = 1.94, p = .163). moral disengagement. The disengagement of moral standards and
Therefore, we did not find convincing evidence that effect sizes self-sanctions can occur, for example, when justifications make
varied across disciplines. Publication bias analyses revealed that it is
the conduct looks less immoral; consequences of the dishonest act
unlikely (table S14). are minimized, ignored, or misconstrued; responsibility for the
dishonest act is displaced or diffused; or victims of the wrongdoing
Moral Reminders are devalued or blamed (Bandura 1999). Similarly, self-concept
Self-concept maintenance theory suggests maintenance theory (Mazar, Amir, and
that reminding individuals about their moral Self-concept maintenance theory Ariely 2008) argues that the availability of
standards may be an effective way to reduce justifications for dishonest conduct leads
unethical behavior. When reminded about
suggests that reminding individ- to increased unethical behavior by allowing
their moral standards, individuals cannot uals about their moral stand- individuals not to update their self-image
get away with behaving unethically without ards may be an effective way to as honest and good. The mechanism of
updating their self-concept (Mazar, Amir, reduce unethical behavior. self-justifications seems to hold regardless
and Ariely 2008; Shu et al. 2012). Focus of the fact that the justification is relevant
theory of normative conduct suggests that or irrelevant for the unethical choice at
injunctive messages set behavioral expectations (Cialdini 2007; hand (Hsee 1996; Schweitzer and Hsee 2002). We identified 13
Cialdini et al. 2006). Codes of ethics can communicate injunctive independent experiments (n = 1,195) testing whether the availability
messages by detailing behaviors that will be judged acceptable versus of self-justifications leads to greater unethical behavior.
nonacceptable (Bing et al. 2012; Gino, Ayal, and Ariely 2009). We
identified 15 experiments (n = 1,492) that tested the effects of moral Meta-analytic findings showed a significant effect of the availability
reminders on individuals’ unethicality. Experiments in this meta- of self-justifications on unethicality. The Hedges’s g was .88
What Causes Unethical Behavior? A Meta-Analysis to Set an Agenda for Public Administration Research 333
(p < .001) (table S6). The heterogeneity test indicated significant Discussion
variability among studies (Q-value (12) = 70.89 (p < .001); I2 = 83.07 We conducted a comprehensive meta-analysis of 137 independent
percent; T2 = .214). We conducted a subgroup analysis to test for experiments from 73 articles to identify the determinants of
heterogeneity across disciplines. Hedges’s g point estimates of unethical behavior. Our results (table 1) supported the previous
the effect sizes were .86 (p < .001, N = 3) for economics and .92 literature, suggesting that specific types of social influences
(p < .001, N = 9) for psychology (Q-value = .04, p = .840). Therefore, (such as examples of unethical behavior by in-group members
we found no evidence of heterogeneity of summary effect size across or interdependence with other individuals benefiting from our
disciplines. unethical actions), greed, egocentrism, self-justifications, exposure
to incremental dishonesty (slippery-slope effect), loss aversion,
Furthermore, we ran a subgroup analysis by observed outcomes. challenging performance goals, and time pressure increase unethical
Hedges’s g point estimates of the effect sizes were .92 (p < .001, behavior. Moreover, in line with theoretical predictions based on
N = 10) for actual behavior and .77 (p < .001, N = 3) for intentions the research, our meta-analytic findings revealed that monitoring of
to behave (Q-value = .34, p = .56). Thus, effect sizes did not vary by employees, providing moral reminders (e.g., exposing subjects to a
outcomes measured in primary studies. Publication bias is unlikely code of ethics), and individuals’ willingness to maintain a positive
in this case (table S14). self-view decrease unethical conduct. We did not find clear evidence
that self-control depletion increases unethical behavior.
Self-View
Self-concept maintenance theory suggests that individuals strive Heterogeneity measures (Q-value, I2, T2) indicated significant
to create and maintain an image of themselves as good and ethical variability among primary data for the following factors: social
persons. On the other side, though, it recognizes that dishonesty influences, greed, monitoring, egocentrism, and self-justification.
pays and individuals are not immune to the temptation of Significant levels of heterogeneity are frequent in meta-analyses of
benefiting from misconduct. To solve the tension, individuals tend social science research (e.g., Cantarelli, Belardinelli, and Belle 2015;
to engage in unethical behavior only up to the point at which they Kish-Gephart, Harrison, and Treviño 2010) and sometimes are used
are not forced to weaken their self-concept (Mazar, Amir, and to justify the preference for random-effect models (e.g., Gerrish
Ariely 2008). The theories of de-individuation (Zimbardo 1969) 2016).
and objective self-awareness (Duval and Wicklund 1972) suggest
that individuals who are more aware of the link between themselves The nature of our primary data prevented us from running meta-
and their actions are less capable of engaging in unethical behavior regressions, which are commonly used to test the moderating effect
without updating their self-image. We identified 12 independent of primary studies’ features such as discipline, publication status,
experiments (n = 859) testing the effects of higher concerns for an experimental design, and type of participants (e.g., Gerrish 2016;
honest and good self-view on individuals’ unethical behavior. Homberg, McCarthy, and Tabvuma 2015). Of the 137 experiments
included in our meta-analyses, about 97 percent were published in
Meta-analytic results showed a significant effect of concerns for self- peer-reviewed journals; 63 percent were conducted in the laboratory
view on unethicality. The Hedges’s g was –.56 (p < .001) (table S7). and 12 percent online (for about 23 percent of the experiments
The heterogeneity test indicated that the variability among studies in our final sample, we could not specify the experimental setting
is not significant at the .05 level (Q-value (11) = 10.18 (p = .515); due to the lack of enough details in the primary study); and 77
I2 = 0.00 percent; T2 = 0.00). We conducted a subgroup analysis to percent used samples of students (table S15). Furthermore, the
test for heterogeneity of summary effect sizes across disciplines. number of experiments in each meta-analysis did not provide the
Hedges’s g point estimates of the effect sizes were –.48 (p = .010, degrees of freedom recommended in the literature to run meta-
N = 2) for business and –.58 (p < .001, N = 10) for psychology regressions (Borenstein et al. 2009). For instance, while the articles
(Q-value = .21, p = .648). Thus, no evidence of heterogeneity of in our meta-analysis were published in several disciplines, including
summary effect size is apparent across disciplines. business, ethics, management, economics, and psychology (table
S15), the sample size of each of our meta-analyses did not allow
Next, we conducted a subgroup analysis to test for heterogeneity us to run meta-regression with discipline as a covariate. However,
across experimental settings, Hedges’s g point estimates of the we conducted subgroup analyses within the factors influencing
effect sizes were –.64 (p < .001, N = 7) for experiments conducted in unethical behavior considering discipline, types of experimental
laboratory settings and –.45 (p < .001, N = 4) for online experiments setting, types of participant, types of outcome observed in primary
(Q-value = 1.67, p = .197). Therefore, we failed to reject the null studies, and sources as eligible grouping variables. Out of all the
hypothesis that effect sizes were the same across experimental subgroup analyses that we ran, we found evidence of significant
settings. In other words, we found no evidence of significant heterogeneity among summary effect sizes at the .05 level in two
differences between laboratory and online studies. cases. Specifically, effect sizes varied across types of participants for
the meta-analysis on egocentrism and moral reminders. Following
Lastly, we ran subgroup analyses by types of participants. Hedges’s recent recommendations and practice (e.g., Cantarelli, Belardinelli,
g point estimates of the effect sizes were –.45 (p < .001, N = 4) and Belle 2016; Kepes et al. 2012), publication bias plots (table
for the general population, –.68 (p < .001, N = 6) for students, S13) and indices (table S14) were computed for the meta-analyses
and –.52 (p = .006, N = 2) for students and university employees based on at least 10 primary experiments. We found no evidence
(Q-value = 2.18, p = .337). Thus, we found no evidence that effect of publication bias for four factors (egocentrism, moral reminders,
sizes differed by type of participants. Lastly, we did not find any sign self-justification, and self-view) and mixed evidence for three factors
of publication bias (table S14). (social influences, greed, and monitoring).
334 Public Administration Review • May | June 2017
The evidence from the experiments included in our meta-analyses Ethics Committee (ASPA 2015). Similarly, the International
apply equally to public and private settings. In fact, 134 out of City/County Management Association’s Code of Ethics aims at
137 primary studies were designed to investigate the behavioral “providing principles of conduct to members” (Eskridge, French,
mechanisms that drive unethical behavior across human relations, and McThomas 2012, 127). Svara (2014) recently summarized
situations, time, and without reference to any specific jobs, the historical development of arguments in favor and against the
professions, and types of organization. In the two experiments adoption of a code of ethics for public administration; discussed
by Brief et al. (1996), participants were asked to play the role of code of ethics’ usefulness, content, and enforcement; and made
a top-level executive making decisions about financial reporting. suggestions about how ASPA can help civil servants address ethical
Ponemon (1992) used a sample of auditors of a public accounting challenges. Indeed, our meta-analytic findings leaves unanswered
firm undergoing training. We acknowledge that results from questions about how effective codes of ethics serve as moral
laboratory experiments may raise concerns about the generalizability reminders. For example, future work should investigate whether
of findings to real public administration settings. However, existing a code needs to emphasize positive or negative behaviors and
evidence indicates the external validity of experiments employing consequences or both or how employees can be best reminded of
samples of students in laboratories may be stronger than expected ethics in the context of implementing and using the codes.
(e.g., List and Levitt 2005).
Some of the additional analyses whose details are not included in
Implications for Public Administration Research and the printed version of the manuscript due to length constraints also
Practice provide relevant insights for public administration. For instance,
Among the factors that our meta-analyses our meta-analytic evidence indicates that
identified as determinants of unethical goal setting and time pressure might increase
behavior, the following may be elevated
Our meta-analytic evidence unethical behavior. These findings raise
to prominence for public administration indicates that goal setting and potential concerns about the unintended
research and practice. First, results from the time pressure might increase consequences of performance management
meta-analyses on social influences suggest unethical behavior. reforms on public service providers’ ethical
that being exposed to corrupted colleagues conduct. In fact, commitment to reach
may enhance the likelihood that one engages unfeasible goals may well enhance the
in unethical conduct. These findings are particularly relevant likelihood that employees use task strategies to achieve them that
because “[c]orruption in the public sector hampers the efficiency of disregard ethical standards (Grant and Shin 2012). Indeed, Bevan
public services, undermines confidence in public institutions and and Hood (2006), Le Grand (2010), and Hood (2012) show that
increases the cost of public transactions” (OECD 2015). Moreover, the adoption of a target and terror regime may lead public servants
corruption “may distort government’s public resource allocations” to cut corners and game performance measurement systems.
(Liu and Mikesell 2014, 346). Relatedly, Lavertu (2016) argues that pressures to increase the
availability of performance data for public organization may lead to
Second, meta-analytic evidence that monitoring reduces unethical misperceptions and biased decisions by policy makers and citizens.
behavior provides support for government initiatives around the
globe aimed at increasing monitoring of public service providers At the aggregated level, the majority of the factors that we identified
through transparency and openness. For instance, U.S. president through the extensive literature review as encouraging or reducing
Barack Obama introduced his open government initiatives unethical behavior align nicely with Kohlberg’s (1980) six stages of
by saying, “My Administration is committed to creating an moral development. Monitoring refers to Kohlberg’s first stage of
unprecedented level of openness in Government” (White House moral development, that is, obedience. Greed, egocentrism, self-
2009). Indeed, we agree with an anonymous reviewer that this can justification, self-view, and goal setting pertain to Kohlberg’s stage
have a potentially broader implication for theory and practice in two, instrumental egoism. Social influences fall under Kohlberg’s
public administration. Namely, the search for institutional designs stage three, interpersonal accord and conformity. Moral reminders
for accountability in which the inclusion of monitoring mechanisms align with Kohlberg’s stage four, duty and social order. Thus, public
into structures could affect ethical behavior. organizations interested in making good use of Kohlberg’s work to
assess and promote the moral development of their employees can
Third, one of our meta-analyses showed that moral reminders can also capitalize on the findings of our meta-analyses.
be a useful tool to prevent and reduce unethical behavior in the
workplace. This finding speaks to the debate on the effectiveness Advancing Research and Practice in Public
of codes of ethics in promoting ethical competence (e.g., Meine Administration
and Dunn 2013; Svara 2014) and supports the introduction Our comprehensive meta-analyses revealed several gaps in the
of codes of ethics for government service worldwide. Examples experimental research on the causes of unethical behavior, thus
include the Code of Conduct for Commissioners adopted by the pointing to directions that public administration scholars could
European Commission and the Standards of Ethical Conduct for take to contribute to the advancement of knowledge in this area.
Employees of the Executive Branch published in 1992 by the U.S. First, to the best of our knowledge, no experimental study has been
Office of Government Ethics. Professional organizations have also published on the causal effects of ethical leadership on followers’
been active in promoting high standards of ethical conduct among unethical behavior. Joining recent calls to address this issue (e.g.,
practitioners in the public sector. For instance, ASPA recently Hassan, Wright, and Yukl 2014), we urge public administration
revised its Code of Ethics for public servants and established an scholars to spearhead experimental work in this area. Research on
What Causes Unethical Behavior? A Meta-Analysis to Set an Agenda for Public Administration Research 335
ethical leadership is flourishing in our field (e.g., Downe, Cowell to always describe the distribution of students by their major area
and Morgan 2016; Hassan, Wright, and Yukl 2014) and in the of study. In addition to field experiments, public administration
social sciences (e.g., Yukl et al. 2013). However, most of these scholars should take full advantage of quasi-experimental
studies rely on correlational designs that are subject to questions methodologies. For instance, when an intervention (e.g., the
about the internal validity of the findings. The use of randomized adoption of a new code of conduct) is introduced in only one of
experiments might help validate the results of extant observational several regions, areas, districts, or organizations that are otherwise
work showing that ethical leadership promotes ethical conduct similar, difference-in-differences is superior to correlational
among followers (e.g., Hassan, Wright, and Yukl 2014). We studies when it comes to internal validity. Public sector reforms
acknowledge that conducting experimental research on unethical often provide naturally occurring experiments that can be wisely
behavior may be a daunting effort (e.g., Hassan, Wright, and exploited to investigate the causes of unethical behavior in the
Yukl 2014). Nevertheless, the richness of our data set, which workplace. As an example, the devolution of powers involving the
includes peer-reviewed studies published in the top journals of delivery of public services that occurred in the 1990s in the United
several disciplines (e.g., Academy of Management Journal, Journal of Kingdom provides a natural experiment with which to compare the
Applied Psychology, Journal of the European Economic Association), likelihood that civil servants would try to game the system under
demonstrates that experimental work in this area is possible different public service delivery models (Bevan and Hood 2006; Le
and widely accepted. The parallel advancement in observational Grand 2010). Regression discontinuity designs might be effective
research (e.g., Bedi, Alpaslan, and Green 2015) can make up for the to understand the determinants of unethicality in the context of
limitations that are inherent to artificial manipulations of ethical naturally occurring experiments.
conduct in an experimental setting (Wright and Grant 2010). As a
general rule, it seems always desirable to triangulate findings using Conclusion
different methodologies. In line with previous literature, our meta-analyses of 137
experiments show that specific types of social influences, greed,
Second, as suggested by one reviewer, public administration egocentrism, self-justifications, exposure to incremental dishonesty,
scholars should explore how the interface between political and loss aversion, challenging performance goals, and time pressure,
career employees may produce unethical behavior. For instance, do increase unethical behavior. On the opposite, monitoring of
corrupt politicians help breed corruption among others, elected and employees, providing moral reminders, and individuals’ willingness
nonelected, by their example? In general, is the ethical behavior of to maintain a positive self-view decrease unethical conduct.
public servants influenced by exemplars—good and bad—to whom
they are exposed in their work? While a group of studies in our Our systematic review points to ways in which scholars can
meta-analysis tested the role of bad examples on other individuals’ advance understanding of ethics in public administration. In
unethicality (e.g., Paternoster et al. 2013), none explored the particular, we urge researchers in our field to use experimental
influence of ethical examples on an individual’s conduct. Therefore, methodologies to address a number of unanswered questions.
studying the effects of being exposed to good examples (whether For instance, future work should explore the determinants of
from leaders or not) maybe particularly interesting from a unethical behavior at the nexus between politicians and career
theoretical perspective. As an initial step to fill this gap, scholars can employees. Moreover, we encourage public administration
replicate the numerous studies that manipulate unethical behavior scholars to test whether being exposed to good examples and
by reversing the experimental interventions. ethical leaders may affect civil servants’ unethicality. Finally, our
results suggest that unethical behavior by street-level bureaucrats
Third, research on empathy and altruism suggests that the desire to who empathize with their clients is another area that deserves
help others may trigger unethical behavior, in particular when the research attention.
beneficiary is identified (e.g., Batson et al. 1991). This finding has
potential implications for the study of unethical behavior by street- Acknowledgments
level bureaucrats and case managers who might be tempted to bend We are grateful to Paul Battaglio, Doug Goodman, L. Douglas
the rules to help clients to whom they are emotionally bound (e.g., Kiel, Young-joo Lee, and Jim Perry for their comments on previous
Bozeman and Su 2015; Perry and Vandenabeele 2015). versions of this work.
What Causes Unethical Behavior? A Meta-Analysis to Set an Agenda for Public Administration Research 337
and Willingness to Report Ethical Problems. Public Administration Review 74(3): Messick, David M., and Charles G. McClintock. 1968. Motivational Bases of
333–43. Choice in Experimental Games. Journal of Experimental Social Psychology 4(1):
Higgins, Julian P. T., and Sally Green. 2011. Cochrane Handbook for Systematic 1–25.
Reviews of Interventions, version 5.1.0. http://www.handbook.cochrane.org Milgram, Stanley. 1963. Behavioral Study of Obedience. Journal of Abnormal and
[accessed September 3, 2016]. Social Psychology 67(4): 371–78.
Homberg, Fabian, Dermot McCarthy, and Vurain Tabvuma. 2015. A Meta-Analysis Moore, Celia, and Francesca Gino. 2015. Approach, Ability, Aftermath: A
of the Relationship between Public Service Motivation and Job Satisfaction. Psychological Process Framework of Unethical Behavior at Work. Academy of
Public Administration Review 75(5): 711–22. Management Annals 9(1): 235–89.
Hood, Christopher. 2012. Public Management by Numbers as a Performance- Organisation for Economic Co-operation and Development (OECD). 2015.
Enhancing Drug: Two Hypotheses. Special issue, Public Administration Review Fighting Corruption in the Public Sector. http://www.oecd.org/corruption/
72: S85–92. ethics/ [accessed December 14, 2016].
Houser, Daniel, Stefan Vetter, and Joachim Winter. 2012. Fairness and Cheating. Paternoster, Ray, Jean Marie McGloin, Holly Nguyen, and Kyle J. Thomas.
European Economic Review 56(8): 1645–55. 2013. The Causal Impact of Exposure to Deviant Peers: An Experimental
Hsee, Christopher K. 1996. Elastic Justification: How Unjustifiable Factors Influence Investigation. Journal of Research in Crime and Delinquency 50(4): 476–03.
Judgments. Organizational Behavior and Human Decision Processes 66(1): 122–29. Perry, James L. 2012. How Can We Improve Our Science to Generate More Usable
Insko, Chester A., John Schopler, Rick H. Hoyle, Gregory J. Dardis, and Kenneth A. Knowledge for Public Professionals? Public Administration Review 72(4):
Graetz. 1990. Individual-Group Discontinuity as a Function of Fear and Greed. 479–82.
Journal of Personality and Social Psychology 58(1): 68–79. ———. 2015. Revisiting the Core of Our Good Government Ethos. Public
John, Leslie K., George Loewenstein, and Scott I. Rick. 2014. Cheating More for Administration Review 75(2): 186–87.
Less: Upward Social Comparisons Motivate the Poorly Compensated to Cheat. Perry, James L., and Wouter Vandenabeele. 2015. Public Service Motivation
Organizational Behavior and Human Decision Processes 123(2): 101–9. Research: Achievements, Challenges, and Future Directions. Public
Kepes, Sven, George C. Banks, Michael McDaniel, and Deborah L. Whetzel. 2012. Administration Review 75(5): 692–99.
Publication Bias in the Organizational Sciences. Organizational Research Methods Ploner, Matteo, and Tobias Regner. 2013. Self-Image and Moral Balancing: An
15(4): 624–62. Experimental Analysis. Journal of Economic Behavior & Organization 93:
Kish-Gephart, Jennifer J., David A. Harrison, and Linda Klebe Treviño. 2010. Bad 374–83.
Apples, Bad Cases, and Bad Barrels: Meta-Analytic Evidence about Sources of Ponemon, Lawrence A. 1992. Auditor Underreporting of Time and Moral Reasoning:
Unethical Decisions at Work. Journal of Applied Psychology 95(1): 1–31. An Experimental Lab Study. Contemporary Accounting Research 9(1): 171–89.
Kogut, Tehila, and Ilana Ritov. 2005. The “Identified Victim” Effect: An Identified Pugh, Darrell. 1991. Origins of Ethical Frameworks in Public Administration. In
Group, or Just a Single Individual? Journal of Behavioral Decision Making 18(3): Ethical Frontiers in Public Management, edited by James S. Bowman, 9–32. San
157–67. Francisco: Jossey-Bass.
Kohlberg, Lawrence. 1980. The Meaning and Measurement of Moral Development. Rainey, Hal G. 2009. Understanding and Managing Public Organizations, 4th ed. San
Worcester, MA: Clark University Press. Francisco: Jossey-Bass.
Lavertu, Stéphane. 2016. We All Need Help: Big Data and the Mismeasure of Public Rest, James R., Darcia Narvaez, Muriel J. Bebeau, and Stephen J. Thoma. 1999.
Administration. Public Administration Review 76(6): 864–72. Postconventional Moral Thinking: A Neo-Kohlbergian Approach. Mahwah, NJ:
Le Grand, Julian. 2010. Knights and Knaves Return: Public Service Motivation and Lawrence Erlbaum.
the Delivery of Public Services. International Public Management Journal 13(1): Rigdon, Mary L., and Alexander P. D’Esterre. 2015. The Effects of Competition on
56–71. the Nature of Cheating Behavior. Southern Economic Journal 81(4): 1012–24.
Lim Choi, Do. 2006. Moral Reasoning in Public Service: Individual, Organizational, and Ringquist, Evan J. 2013. Meta-Analysis for Public Management and Policy. San
Societal Determinants. International Review of Public Administration 11(1): 35–45. Francisco: Jossey-Bass.
List, John A., and Steven D. Levitt. 2005. What Do Laboratory Experiments Tell Rixom, Jessica, and Himanshu Mishra. 2014. Ethical Ends: Effect of Abstract
Us about the Real World? Working paper. Cambridge: National Bureau for Mindsets in Ethical Decisions for the Greater Social Good. Organizational
Economic Research. Behavior and Human Decision Processes 124(2): 110–21.
Liu, Cheol, and John L. Mikesell. 2014. The Impact of Public Officials’ Corruption Rizzo, Ann-Marie, and Laura Lee Swisher. 2004. Comparing the Stewart–Sprinthall
on the Size and Allocation of U.S. State Spending. Public Administration Review Management Survey and the Defining Issues Test as Measures of Moral
74(3): 346–59. Reasoning in Public Administration. Journal of Public Administration Research
Mazar, Nina, and Pankaj Aggarwal. 2011. Greasing the Palm: Can Collectivism and Theory 14(3): 335–48.
Promote Bribery? Psychological Science 22(7): 843–48. Rohr, John A. 1986. To Run a Constitution: The Legitimacy of the Administrative State.
Mazar, Nina, On Amir, and Dan Ariely. 2008. The Dishonesty of Honest People: Lawrence: University Press of Kansas.
A Theory of Self-Concept Maintenance. Journal of Marketing Research 45(6): ———. 1998. Public Service, Ethics, and Constitutional Practice. Lawrence:
633–44. University Press of Kansas.
Meine, Manfred F., and Thomas P. Dunn. 2013. The Search for Ethical Competency: Rubin, Mark, and Miles Hewstone. 1998. Social Identity Theory’s Self-Esteem
Do Ethics Codes Matter? Public Integrity 15(2): 149–66. Hypothesis: A Review and Some Suggestions for Clarification. Personality and
Menzel, Donald C. 2005. Research on Ethics and Integrity in Governance: A Review Social Psychology Review 2(1): 40–62.
and Assessment. Public Integrity 7(2): 147–68. Sah, Sunita, and George Loewenstein. 2015. Conflicted Advice and Second
———. 2015. Research on Ethics and Integrity in Public Administration: Moving Opinions: Benefits, but Unintended Consequences. Organizational Behavior and
Forward, Looking Back. Public Integrity 1(4): 343–70. Human Decision Processes 130: 89–107.
Menzel, Donald C., and Kathleen J. Carson. 1999. A Review and Assessment of Schweitzer, Maurice E., and Christopher K. Hsee. 2002. Stretching the Truth: Elastic
Empirical Research on Public Administration Ethics: Implications for Scholars Justification and Motivated Communication of Uncertain Information. Journal
and Managers. Public Integrity 1(3): 239–64. of Risk and Uncertainty 25(2): 185–201.
Supporting Information
Supplementary material may be found in the online version of this article at http://onlinelibrary.wiley.com/doi/10.1111/puar.12714/full.
What Causes Unethical Behavior? A Meta-Analysis to Set an Agenda for Public Administration Research 339